With the growth of software systems, logs have become an important data to aid system maintenance. Log-based anomaly detection is one most methods for such purpose, which aims automatically detect anomalies via log analysis. However, existing log-based approaches still suffer from practical issues due either depending on a large amount manually labeled training (supervised approaches) or unsatisfactory performance without learning knowledge historical (unsupervised and semi-supervised...

Deep learning (DL) techniques are rapidly developed and have been widely adopted in practice. However, similar to traditional software systems, DL systems also contain bugs, which could cause serious impacts especially safety-critical domains. Recently, many research approaches focused on testing models, while little attention has paid for libraries, is the basis of building models directly affects behavior systems. In this work, we propose a novel approach, LEMON, libraries. particular, (1)...

Deep Neural Network (DNN) testing is one of the most widely-used ways to guarantee quality DNNs. However, labeling test inputs check correctness DNN prediction very costly, which could largely affect efficiency testing, even whole process development. To relieve labeling-cost problem, we propose a novel input prioritization approach (called PRIMA) for DNNs via intelligent mutation analysis in order label more bug-revealing earlier limited time, facilitates improve testing. PRIMA based on key...

Deep neural network (DNN) has become increasingly popular and DNN testing is very critical to guarantee the correctness of DNN, i.e., accuracy in this work. However, suffers from a serious efficiency problem, it costly label each test input know for set, since labeling involves multiple persons (even with domain-specific knowledge) manual way set large-scale. To relieve we propose novel practical approach, called PACE (which short P ractical AC curacy E stimation), which selects small inputs...

Pre-training and fine-tuning have emerged as a promising paradigm across various natural language processing (NLP) tasks. The effectiveness of pretrained large models (LLM) has witnessed further enhancement, holding potential for applications in the field medicine, particularly context Traditional Chinese Medicine (TCM). However, application these general to specific domains often yields suboptimal results, primarily due challenges like lack domain knowledge, unique objectives, computational...

Java Virtual Machine (JVM) provides the runtime environment for programs, which allows to be "write once, run anywhere". JVM plays a decisive role in correctness of all programs running on it. Therefore, ensuring and robustness implementations is essential programs. To date, various techniques have been proposed expose bugs via generating potential bug-revealing test However, diversity effectiveness generated by existing research are far from enough since they mainly focus minor...

We present a novel defending strategy, adaptive Markov strategy (AMS), to protect smart-grid system from being attacked by unknown attackers with unpredictable and dynamic behaviors. One significant merit of deploying AMS defend the is that it theoretically guaranteed converge best response against any stationary attacker, Nash equilibrium (NE) in case self-play (the attacker intelligent enough use attack). The effectiveness evaluated considering class data integrity attacks which an manages...

PLELog is a novel approach for log-based anomaly detection via probabilistic label estimation. It designed to effectively detect anomalies in unlabeled logs and meanwhile avoid the manual labeling effort training data generation. We use semantic information within log events as fixed-length vectors apply HDBSCAN automatically clustering sequences. After that, we also propose Probabilistic Label Estimation reduce noises introduced by error put "labeled" instances into an attention-based GRU...

Numerical computation is dominant in deep learning (DL) programs. Consequently, numerical bugs are one of the most prominent kinds defects DL can lead to exceptional values such as NaN (Not-a-Number) and INF (Infinite), which be propagated eventually cause crashes or invalid outputs. They occur when special inputs parameter at internal mathematical operations log(). In this paper, we propose first dynamic technique, called GRIST, automatically generates a small input that expose GRIST...

Deep learning (DL) systems have been widely utilized across various domains. However, the evolution of DL can result in regression faults. In addition to through incorporation new data, feature evolution, such as features, is also common and introduce this work, we first investigate underlying factors that are correlated with faults scenarios, i.e., redundancy contribution shift. Based on our investigation, propose a novel mitigation approach called FeaProtect, which aims minimize impact...

Developing convincing and realistic virtual human behavior is essential for enhancing user experiences in reality (VR) augmented (AR) settings. This paper introduces a novel task focused on generating long-term behaviors agents, guided by specific personality traits contextual elements within 3D environments. We present comprehensive framework capable of autonomously producing daily activities autoregressively. By modeling the intricate connections between characteristics observable...

Security vulnerability prediction (SVP) can identify potential vulnerable modules in advance and then help developers to allocate most of the test resources these modules. To evaluate performance different SVP methods, we should take security audit code inspection into account consider effort-aware measures (such as ACC P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">opt</sub> ). However, best our knowledge, effectiveness methods has not been...

Learning to generate diverse scene-aware and goal-oriented human motions in 3D scenes remains challenging due the mediocre characteristics of existing datasets on Human-Scene Interaction (HSI); they only have limited scale/quality lack semantics. To fill gap, we propose a large-scale semantic-rich synthetic HSI dataset, denoted as HUMANISE, by aligning captured motion sequences with various indoor scenes. We automatically annotate aligned language descriptions that depict action unique...

Deep learning (DL) Systems have been widely used in various domains. Similar to traditional software, DL system evolution may also incur regression faults. To find the faults between versions of a system, we propose novel fuzzing technique called DRFuzz, which facilitates generating inputs that trigger diverse and high fidelity. enhance diversity found faults, DRFuzz proposes diversity-oriented test criterion explore as many faulty behaviors possible. Then, incorporates GAN model guarantee...

The broker mechanism is widely applied to serve for interested parties derive long-term policies in order reduce costs or gain profits smart grid. However, a faced with number of challenging problems such as balancing demand and supply from customers competing other coexisting brokers maximize its profit. In this paper, we develop an effective pricing strategy local electricity retail market based on recurrent deep multiagent reinforcement learning sequential clustering. We use real...

Clone detection techniques have been explored for decades. Recently, deep learning has adopted to improve the code representation capability, and state-of-the-art in clone detection. These approaches usually require a transformation from AST binary tree incorporate syntactical information, which introduces overheads. Moreover, these conduct term-embedding, requires large training datasets. In this paper, we introduce embedding technique Our approach first conducts obtain node vector each...

The global coronavirus disease 2019 (COVID-19) pandemic seriously affected people's lives. We evaluated anxiety and depression among patients with insomnia in northeast China during the first wave release of COVID-19, providing a basis for clinical diagnosis treatment insomnia.

Due to the critical role of compilers, many compiler testing techniques have been proposed, two most notable categories among which are grammar-based and metamorphic-based techniques. All them extensively studied for mature compilers. However, it is typical develop a new new-born programming language in practice. In this scenario, existing hardly applicable due some major reasons: (1) no reference compilers support differential testing, (2) lack program analysis tools (3) substantial...

For programmers, learning the usage of APIs (Application Programming Interfaces) a software library is important yet difficult. API recommendation tools can help developers use by recommending which to be used next given that have been written. Traditionally, language models such as N-gram are applied recommendation. However, because libraries keep changing and new emerging, common. These seen OOV (out vocabulary) words cannot handled well existing approaches due lack training data. In this...